2
\$\begingroup\$

The platform is having various supply voltages generated by the power supply: +5V (for digital part), +12V and -12V for analog part, including video amplifiers.

My device, which is being installed into the platform, has several switching power supplies based on Maxim step down converters. One of them, converting +5V to +3.3V, powers the DAC.

I designed the audio circuit the following way: DAC output, through coupling capacitor, goes to the op-amps, which are powered from +-12V of the platform. I can not power op-amps from the 3.3V source because their inputs do also get negative voltage (due to coupling cap, and it can not be changed).

The problem

Due to platform powering video circuitry, +-12V power rails are badly affected by the noise from video amplifiers, and if DAC is silent, then output of op-amps, if amplified at about 50 dB, sound strong beeping and bleeping sound in the audible frequency range. If I disconnect DAC output from the op-amp circuitry, and leave only pull-down 1MOhm resistor to the ground at the op-amp input, this beeping/bleeping noise becomes much more loud and can be heard without amplification (because op-amp coupling with ground becomes weaker than when DAC is connected).

I checked DAC output (being powered from 3.3V generated by 5V -> 3.3V converter), and, when DAC digital input is constant, DAC generates white noise at its output at much much less magnitude, thus bleeping does not come from DAC but definitely from +-12V power rails.

What I have tried

  1. Passive filtering of +-12V power. No help, and I believe it is useless because bleeping noise depends on the image being displayed. No change to picture = silence or low-patterned noise, text scroll = massive noise in audible spectrum.

  2. Passive filtering with TPS7A49/TPS7A30 from TI, performing 12V -> 10V conversion on both power rails. No help, noise is still in there, and became even worse (I suspect it became worse because I assembled the circuit on the breadboard in dead-bug manner).

  3. IMHO, next step is trying PWM converter, converting +-12V to (let's say) +-9V, or use step-up converter making these voltages from +5V - using high frequency switcher (e.g. 1 MHz) to eliminate problems in the audible frequency domain. I did not find suitable converters from TI, Maxim has MAX15462 (step down), but does not have pair for negative power rail to keep common ground. Another device is MAX8614 (step up using +5V as input), which seems to be within the spec for my current requirements. And I learnt that I need PWM device rather than PFM for the audio, because PFM may easily drop its frequency down to audible band when amplification circuits are "idle" (consuming ~8 mA).

Do I move in correct direction? Is it an overkill? Can you advise any other ways?


Update: special thanks for bobflux for his advice. Powered the op-amp circuit on the breadboard using the 4 x 18650 cells, connected input to the output of DAC and to the ground nearby this output. Noise is still there.

Then recorded output of the input of the op-amp on the device board, which has 1 MOhm pull down, and air-wire soldered to it (which was desoldered from the DAC output). Strong bleeping noise. Connected this wire to the ground (shorting input of the op-amp to the ground), and bleeping noise, while still there audible if amplifying by 80 dB (!), is now at the same level as white noise.

Thus you were right, the op-amp picks up noise, because DAC output and op-amp input are relatively high impedance. Look at this post (it will also give you an idea about layout). I was advised to increase DAC output resistor from 100 Ohms to 1 kOhm, and it seems increased the noise pickup level. Next, I use unshielded wires between this resistor after DAC and op-amp input, wire length is approximately 8 cm (3").

Thus the major problem lies not in the noisy power lines, but in noise pickup (what a pity I can not see or feel this noise around! :). Unfortunately I can not move DAC closer to the op-amps, but I can choose wire types I use and ground connection type.

I can use shielded wires from DAC to op-amps, but what to do with grounding of the shield? If I ground at both ends there will be a ground loop. I have read somewhere that I must ground only at one end, but that was about two different devices on the long transmission line, here I have single device with GND polygon pieces connected together on the board from both sides, thus maybe connecting shielded wires' ground at both ends - near DAC output and near op-amp input/power may be a way to go.

Next, what to do with DAC output impedance, as if I will return back to 100 Ohm resistor, as advised on the typical circuit diagram for UDA1334, there may be excessive input current for op-amps during power off. That would be super if I can put op-amp just after the DAC output lowering circuit's output impedance before transmission further to op-amps, but I do not have +- supply in that place of the board (and actually very tight on space).


Small update on the op-amp impedance: I was using AD8512 before, it has 5 mA input current limit. Now I use ADA4610, having 10 mA current limit. DAC output circuitry may be as high as +3V (depends on the coupling capacitor charge during power off), thus 300 Ohm series resistor must be fairly ok for the protection of the input.

Your opinion?


Update: I decided to make shortcut, and instead of listening to ground noise just solder shielded wires (as it was planned at the very beginning):

enter image description here

and bleeping noise is gone. Totally gone.

Now there's another problem, but not related to the originally asked: what are the best shielded wires for the purpose (aiwires for consumer audio) and how to wire properly not to short circuit components on the board (these were actually reasons why I was not using shielded wires)...

\$\endgroup\$
11
  • 1
    \$\begingroup\$ These issues could also be due to the layout - how many layers is the PCB? Have you made sure to tack all lines to keep suitable reference planes and return paths? Analogue signals can be extremely sensitive, often requiring extra clearance and possibly shielding traces. \$\endgroup\$ – Puffafish Oct 26 '20 at 8:38
  • \$\begingroup\$ The board is double sided, thus only 2 layers. However op-amps are located away from digital stuff, just at the +-12V power inputs to my device. In my opinion, the fact that op-amps with just pull-down resistor at the input, and noise at the output, plus the pattern of noise (which can only come from +12 V power rails) clearly point to the influence from the outside of my device (through power +-12V rails). My device works at MHz frequencies, and I would be ok with white noise, but when bleeping noise comes with screen change... it is simply disgusting. \$\endgroup\$ – Anonymous Oct 26 '20 at 8:44
  • \$\begingroup\$ huh, interesting. So, if you're in RF, then 50 dB is about the best cross-talk isolation you'll achieve on the same double-sided PCB, not much you can do. But you're mostly for below that in frequency, right? The fact that power filtering doesn't help might mean that we're either not looking at noise transported through the power lines, or noise that is much higher in frequency than we think, and thus passes through your filtering methods (which sound very reasonable). \$\endgroup\$ – Marcus Müller Oct 26 '20 at 9:41
  • 2
    \$\begingroup\$ Before fixing the problem, identify the problem. Let's use a divide and conquer approach. Cut the +12 and -12 power traces to the opamps (leave ground alone) and power the opamps only from a clean supply, for example a bench supply or a couple of 9V batteries. They'll run on 9V just fine instead of 12V. Still got noise? Also can you post layout? \$\endgroup\$ – bobflux Oct 26 '20 at 13:50
  • 1
    \$\begingroup\$ NICE. It looks like audiophile now ;) Now if you disconnect both shields from one side and you get noise, you'll know the problem was ground. If you get no noise, then it was picked up by the trace along the way. Useful if you do a new revision. \$\endgroup\$ – bobflux Oct 26 '20 at 20:58
0
\$\begingroup\$

Your noise can couple over two path into the signal. If the power source of the OP Amp is noisi and the Amp does not have enough PSRR (power supply ripple rejection) the noise on the supply will show in the signal. Or it can couple directly to the signal (specially if you are working with high impedance).

To reduce power supply ripple, place a LDO between OPAMP and Supply. Give it at least 1.5 to double of the minimal headroom. This will reduce noise from the source.

Direct coupling can be reduced by increasing the current flowing on the singal lines or by better placement in the PCB.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.